The Online Target Date Assignment Problem
|
|
- Charlotte Booth
- 5 years ago
- Views:
Transcription
1 The Online Target Date Assignment Problem S. Heinz 1,S.O.Krumke 2,N.Megow 3,J.Rambau 4, A. Tuchscherer 1, and T. Vredeveld 5 1 Konrad-Zuse-Zentrum für Informationstechnik Berlin, Department Optimization, Takustr. 7, Berlin, Germany {heinz, tuchscherer}@zib.de 2 University of Kaiserslautern, Department of Mathematics, P.O. Box 3049, Paul-Ehrlich-Str. 14, Kaiserslautern, Germany krumke@mathematik.uni-kl.de 3 Technische Universität Berlin, Institut für Mathematik, Strasse des 17. Juni 136, Berlin, Germany nmegow@math.tu-berlin.de 4 Universität Bayreuth, Lehrstuhl für Wirtschaftsmathematik, Bayreuth, Germany Joerg.Rambau@uni-bayreuth.de 5 Maastricht University, Department of Quantitative Economics, P.O. Box 616, 6200 MD Maastricht, The Netherlands T.Vredeveld@KE.unimaas.nl Abstract. Many online problems encountered in real-life involve a twostage decision process: upon arrival of a new request, an irrevocable firststage decision (the assignment of a specific resource to the request) must be made immediately, while in a second stage process, certain subinstances (that is, the instances of all requests assigned to a particular resource) can be solved to optimality (offline) later. We introduce the novel concept of an Online Target Date Assignment Problem (OnlineTDAP) as a general framework for online problems with this nature. Requests for the OnlineTDAP become known at certain dates. An online algorithm has to assign a target date to each request, specifying on which date the request should be processed (e. g., an appointment with a customer for a washing machine repair). The cost at a target date is given by the downstream cost, the optimal cost of processing all requests at that date w. r. t. some fixed downstream offline optimization problem (e. g., the cost of an optimal dispatch for service technicians). We provide general competitive algorithms for the Online- TDAP independently of the particular downstream problem, when the overall objective is to minimize either the sum or the maximum of all downstream costs. As the first basic examples, we analyze the competitive ratios of our algorithms for the particular academic downstream problems of bin-packing, nonpreemptive scheduling on identical parallel machines, and routing a traveling salesman. Supported by the DFG Research Center Matheon Mathematics for key technologies in Berlin. T. Erlebach and G. Persiano (Eds.): WAOA 2005, LNCS 3879, pp , c Springer-Verlag Berlin Heidelberg 2006
2 The Online Target Date Assignment Problem Introduction Many real-world online problems exhibit a two-stage structure. In a first stage, an immediate online action has to be taken, while in a second stage certain offline subproblems (which we will refer to as downstream optimization problems) can be solved to optimality offline. In this paper we provide a general framework for online problems of this type, the Online Target Date Assignment Problem (OnlineTDAP). As an illustration, consider the following scenario arising in the dispatching of service technicians. When a customer calls in, requesting a maintenance service for his washing machine, one of the service technicians has to visit the customer at its location and fix the problem. This service can be done within a certain time frame, say within a week. The customer must be given the day (and possibly a more narrow time window) when the technician will arrive, while he is on the phone and without knowledge of future service requests, that is, it must be given online. However, until the promised service day arrives, the decision which service technician to send and in which order the customers should be visited can be safely deferred. In other words, the exact scheduling and routing of service technicians for a fixed day can be done optimally offline at the night before. In this paper, we introduce structures that account for the following dichotomy in many day-to-day resource dispatching problems: First, a resource has to be assigned to a request (e. g., assign a service vehicle to a repair request) and then the processing of all requests assigned to a certain resource can be optimized (find an optimal tour for each service vehicle). The assignment decisions influence the overall cost because they determine the input and thus the optimal costs of the single resource dispatching problems, the downstream optimization problems. Offline, both stages can be integrated to obtain an overall optimal solution, even in many practical applications. However, if for each request the first decision, i.e., the assignment decision, has to be made online, the situation changes: the resulting problem is not offline anymore, but it is neither just the online version of the integrated dispatching problem; it is something in between. In stochastic programming the optimal decisions of a second stage optimization are called a recourse. In a way, in this paper we introduce competitive analysis with recourse. Our object of study can be seen as the most extreme distinction between the online requirement of the first decision and the downstream optimization: We present a model where the first decision has to be made immediately and irrevocably before the next request is revealed (no knowledge about the input), while the downstream optimization can be carried out offline (complete knowledge about the input). The resource that has to be assigned to requests in our main actor, the OnlineTDAP, is a target date, a date at which the service should take place. There are many variants conceivable of this concept: if the current day is allowed as a target date then the downstream optimization becomes an online problem as well, although a large portion of the data is known before the target
3 232 S. Heinz et al. date. It is also possible to relax the online requirement of the assignment decision: all requests on a single day might be collected, and the target dates are chosen and communicated at the end of the day. And there are, of course, variants where resources other than dates have to be assigned online (machines, vehicles) before a single resource offline problem has to be solved. Problems of this type are abundant in reality, and very often the first decision is online. There is, however, almost no theoretical background published on this topic for the case where no stochastic information about future requests is available. And many of the stochastic models, e. g., Markov Decision Processes [5], cannot be solved for practical problem sizes. Therefore we feel that the investigation of the most basic structures in such problems seems adequate. Thus, we get started in this paper by investigating competitive online algorithms for the OnlineTDAP w. r. t. to classical downstream problems. We think that the introduction of the OnlineTDAP will foster various lines of research, e. g., dealing with competitive analysis for OnlineTDAP w. r. t. various other, maybe more sophisticated downstream problems, with variants of the OnlineTDAP itself, but also with decision support methods for variants of the OnlineTDAP outside competitive analysis. Problem description. An instance of the OnlineTDAP consists of a sequence of requests σ = r 1,r 2,... and a downstream problem Π, an offline optimization problem for which arbitrary subsets of σ are feasible inputs. Each request r i has an integral release date t(r i ) and must be assigned immediately and irrevocably to a target date in the time period t(r i )+1,...,t(r i )+δ(r i ), where δ(r i ) is the allowed time for deferring the service of request r i (one week in our service technician scenario), which is also revealed upon arrival of the request. In this paper we consider only the case of uniform deferral times, that is, δ(r i )=δ for all requests r i,where1 δ<+. For an algorithm Alg we denote the particular date to which request r i is assigned by Alg[r i ] {t(r i )+1,...,t(r i )+δ}. A solution for an OnlineTDAP w. r. t. to downstream problem Π is feasible if each request is assigned to a feasible target date, and for each single target date, the corresponding instance of Π is feasible, too. Let σ d be the subset of requests assigned to date d by an online algorithm Alg. The optimal cost of Π on σ d is called downstream cost of Alg at date d, and we denote it by downcost(σ d ). The overall online cost Alg(σ) of an online algorithm Alg is defined as either the sum of the incurred downstream costs over all dates (min-total problems), or the maximum of the incurred downstream costs over all dates (min-max problems). The goal is to find online algorithms whose competitive ratios are as small as possible. An online algorithm Alg is called c-competitive if the cost of Alg is never larger than c times the cost of an optimal offline solution. The competitive ratio of Alg is the infimum over all c 1suchAlg is c-competitive [2]. Our results. The OnlineTDAP provides a general framework for a large class of online problems and gives a novel view on online optimization. We provide
4 The Online Target Date Assignment Problem 233 Table 1. Main bounds on the competitive ratio of best possible deterministic online algorithms for the OnlineTDAP with a certain downstream problem minimizing the total or maximum downstream cost downstream problem lower bound upper bound bin-packing 3/2 2 scheduling 2 2 traveling salesman 2 2 Minimizing the total downstream cost (min-total objective). downstream problem lower bound upper bound bin-packing 2 min{4,δ} scheduling 3/2 3 1/δ traveling salesman 2 2δ 1 Minimizing the maximum downstream cost (min-max objective). general competitive online algorithms for the OnlineTDAP and analyze them in greater detail w. r. t. classical combinatorial downstream problems such as bin-packing [4, SR1], nonpreemptive parallel machine scheduling [4, SS8] and the traveling salesman problem [4, ND22]. The algorithms we propose do not depend on the downstream problem (although the analysis does). We emphasize that the particular downstream problems discussed in this paper should be seen mainly as illustrating examples for the general framework. Concerning standard online investigations on these problems, [3] gives surveys on online bin-packing and scheduling; the online traveling salesman problem has been considered in [1]. Within the OnlineTDAP framework, our results are online algorithms and lower and upper bounds on their performance guarantees, the competitive ratio, obtained by classical competitive analysis for online algorithms (see, e. g. [2]). In Section 2 we present a 2-competitive algorithm for the min-total objective, i. e., the objective to minimize the total cost summed over all target dates. In Section 3 we consider min-max problems for which the objective is to minimize the maximum downstream cost that occurs on a target date. Here, we give a general online assignment algorithm that we prove to be 4-competitive for the OnlineTDAP with the bin-packing downstream problem and which is 3-competitive for the scheduling setting. Our main results are summarized in Table 1. Finally, we observe for both objective functions that special profiles for the downstream problem, as e. g., (un-) bounded number of machines or bins per target date, lead to trivial problems or prevent any deterministic online algorithm from achieving a constant competitive ratio. 2 Minimizing Total Downstream Cost In this section, we consider the OnlineTDAP with the objective to minimize the total downstream cost summed up over all target dates (min-total objective). Particular downstream problems we deal with are bin-packing, scheduling on parallel machines, and the traveling salesman problem. We first present our main competitiveness result which is an online algorithm formulated independently of the downstream problem. Let us say that a target date is used, if a request has been assigned to it.
5 234 S. Heinz et al. Algorithm PackTogetherOrDelay (Ptd) Assign a request r to the earliest date in the feasible range t(r) +1,...,t(r)+δ which is already used. If no used target date is feasible for request r, then assign it to the latest feasible target date, that is, to t(r)+δ. The above algorithm always finds a feasible solution under the assumption that the amount of requests that can be assigned to the same target date is not restricted (we call this the case of unlimited resources). Under this assumption at any moment in time at most one feasible target date is used by Ptd. Theorem 1. Consider the OnlineTDAP w. r. t. downstream problem Π with the min-total objective. Assume that there are unlimited resources in Π and suppose that the following properties hold for any subinstance σ of σ: i. The optimal offline cost for the downstream problem Π is a monotonously increasing function, that is, Opt( σ) Opt(σ) (i. e., Π is monotone). ii. For each disjoint partition σ (1),...,σ (k) of the subsequence σ the inequality downcost( σ) k i=1 downcost(σ(i) ) holds (i. e., Π allows for synergy). Then, algorithm Ptd is 2-competitive. Proof. For a given sequence of requests σ consider the target dates d 1 <d 2 <... < d k that Ptd chooses. Denote by σ odd (and σ even ) the subsequence of requests that the algorithm assigns to target dates d i with odd (respective even) index i. Observe that, if the input to Ptd were solely σ odd or σ even, then each request would still be assigned to the same target date as when operating on σ. Therefore, Ptd(σ) =Ptd(σ odd )+Ptd(σ even ). (1) Moreover, we know by definition of the algorithm that the difference between any two used target dates is at least δ. Thus, the distance between any two different target dates designated for two requests of the subsequence σ odd (or σ even, respectively) is at least 2δ. This implies that no two requests of the same subsequence σ odd (or σ even, respectively) that have not been assigned to the same target date share a single feasible target date. Therefore, no algorithm can assign such two requests to the same target date. With property (ii) we conclude that Ptd(σ odd )=Opt(σ odd ) and Ptd(σ even )=Opt(σ even ). It follows with (1) and the monotonicity condition (i) that we have online cost Ptd(σ) = Opt(σ odd )+Opt(σ even ) 2 Opt(σ). Note, that in the case that property (ii) only holds in a relaxed version with a factor α, i.e.,downcost( σ) α k i downcost(σ(i) ), Ptd is 2α-competitive. We will now demonstrate the power of Theorem 1 by applying it to various instantiations of the OnlineTDAP.
6 2.1 Downstream Bin-Packing The Online Target Date Assignment Problem 235 In bin-packing n items with sizes s 1,...,s n need to be packed in unit sized bins. The objective is to find a packing such that the total size of the items packed in one bin does not exceed the bin s capacity and the total number of bins needed to pack the items is minimized. In OnlineTDAP w. r. t. bin-packing, a request r =(t(r),s(r)) is given by its release date t(r) and its size 0 <s(r) 1. We assume that the number of available bins per day is not bounded because this would disable any deterministic online algorithm to guarantee a feasible solution. The objective is to find an assignment of requests to feasible target dates that minimizes the total sum of used bins of all target dates. The following theorem gives a lower bound on the competitive ratio of any deterministic online algorithm. Theorem 2. No deterministic online algorithm for OnlineTDAP w. r. t. binpacking minimizing the min-total objective has a competitive ratio less than 3/2. Proof. The adversarial sequence starts with a request r 1 released at time 0 with size s(r 1 ) < 1/2. Consider an online algorithm, Alg, that does not assign this request to its deadline δ. ThenattimeAlg[r 1 ] a second request is released with size s(r 2 )=1 s(r 1 ). Alg cannot assign this request to the same date as the first request and therefore it needs two bins, whereas the optimum needs only one. Now consider an online algorithm Alg that assigns the first request to its deadline δ. Thenattimet(r 2 ) = 1 a second request with size s(r 2 )=s(r 1 )is released. If the algorithm does not pack this item with the first request, then it needs two bins and the optimum needs only one. Otherwise, at time t(r 3 )=δ 1 and t(r 4 )=δ two requests are released both with size s(r 3 )=s(r 4 )=1 s(r 1 ). To pack these items, Alg needs two extra bins, thus in total three bins, whereas the optimum would pack request r 1 and r 3 to date δ and item r 2 and r 4 to δ +1, needing only two bins. Since the properties of Theorem 1 are met, we immediately have the following result. Theorem 3. The competitive ratio of Ptd minimizing the total number of used bins for OnlineTDAP w. r. t. bin-packing is 2. That Ptd cannot achieve a better competitive ratio than 2, can be shown by the following instance. For given k N,k 3, let ε<1/(2k 4) and σ = σ (1)... σ (k). σ (1) consists of the following three requests: r 1 =(0, 1), r 2 =(1, 1/2 ε), r 3 =(δ, 1/2+ε). For i =2,...,k, the subsequence σ (i) is defined by σ (i) =((iδ 1, 1/2+(i 2)ε), (iδ, 1/2 (i 2)ε)).
7 236 S. Heinz et al. The cost of Ptd on this sequence is Ptd(σ) =2k + 1. On the other hand, the number of required bins of the optimal offline algorithm is Opt(σ) =k +1.By letting k, the lower bound follows. We conjecture that the following online algorithm, PackFirstOrDelay, has a better performance guarantee than Ptd although the analysis for the general problem seems more difficult. Algorithm PackFirstOrDelay (Pfd) If there is a used target date to which the current request r can be assigned without increasing the number of necessary bins, then the earliest of these dates is chosen. Otherwise, assign the latest possible date, t(r)+δ. This algorithm achieves a better solution on the lower bound instance for Ptd from above. However, there exist instances for which Pfd performs worse than Ptd, as for example: r 1 =(0, 2/5), r 2 =(0, 1/5), r 3 =(0, 1/5), r 4 =(δ 1, 2/5), r 5 =(δ 1, 2/5), and r 6 =(δ 1, 2/5). If all items have identical size the problem becomes much easier. Theorem 4. Consider the OnlineTDAP w. r. t. bin-packing with the min-total objective. Then, Pfd is optimal if all item sizes are equal. Proof. Assume that the bin-packing instance at each date is solved in such a way that at most one bin is partially filled. Given a sequence σ, letpfd(σ) =f + p, where f is the number of full bins and p is the number of partially filled bins. Let d 0 <d 1 <... < d p be the dates on which Pfd has partially filled bins. Let σ be the subsequence consisting of all requests that are packed in a full bin and for each partially filled bin the request that opened this bin. Note that Pfd(σ )=Pfd(σ). We partition σ into subsequences σ l consisting of all requests r σ with d l 1 t(r) <d l.asthelastrequestinσ l and the first request in σ l+1 are both assigned using the delay tactic of Pfd, we know that there is no overlap in the feasible target dates of requests of different subsequences. Hence, Opt(σ )= l Opt(σ l). Moreover, Pfd packs the items of a subsequence in all but one fully filled bins and thus Pfd(σ l )=Opt(σ l ). Combining these equalities, we get Pfd(σ) =Pfd(σ )= l Pfd(σ l )= l Opt(σ l )=Opt(σ ) Opt(σ). 2.2 Downstream Parallel-Machine Scheduling In this section, we consider the OnlineTDAP w. r. t. nonpreemptive machine scheduling of jobs on identical parallel machines to minimize the makespan, i. e., the latest completion time of all jobs on all machines of one date. The overall objective is now to minimize the sum of makespans over all target dates. For convenience we will use standard scheduling terminology, i. e., a request r is a
8 The Online Target Date Assignment Problem 237 job that has a processing time denoted by p(r). We denote a request by an ordered pair of release date and processing time, r =(t(r),p(r)). The number of machines available per date is denoted by m. Note, that in case m =1,the problem is trivial since any target date assignment yields a total downstream cost of r σ p(r), for any sequence σ. Therefore, we assume for the remainder of this section that more than one machine are available each date. Consider the general online algorithm Ptd. Also for this setting with the scheduling downstream problem, Theorem 1 applies and Ptd is 2-competitive. The analysis is tight as the following sequence shows: r 1 =(0,ε), r 2 =(Ptd[r 1 ] 1, 1), r 3 =(Ptd[r 1 ], 1), where ε<1. The costs incurred by the algorithm are Ptd(σ) = 2, whereas optimal offline costs are Opt(σ) = 1 + ε. Thus, we have shown: Theorem 5. The deterministic online algorithm Ptd has a competitive ratio of 2 for the OnlineTDAP for downstream scheduling on identical parallel machines (m >1) subject to minimize the sum of makespans induced on all target dates. Moreover, we obtain the following general lower bound result for this problem setting. Theorem 6. No deterministic online algorithm can achieve a competitive ratio less than 2 for the OnlineTDAP minimizing the total downstream cost caused by nonpreemptive scheduling on more than one machine. Proof. In order to obtain this bound consider for a given online algorithm Alg the following sequence: r 1 =(0, 1), r 2 =(Alg[r 1 ] 1, 1+ 2). If Alg assigns a target date different from Alg[r 1 ]torequestr 2, then no further requests are given. Thus, Alg s cost is Alg(r 1,r 2 )=2+ 2, whereas an offline optimum yields a solution with cost Opt(r 1,r 2 )=1+ 2, which gives a ratio of 2. Assume that Alg assigns request r 2 to the same date as r 1, and a third request r 3 =(Alg[r 1 ], 1+ 2) is given. Then the cost of the online algorithm is , whereas the optimal offline costs are Again, the ratio of the incurred costs of Alg and Opt is 2. Note that the lower bound construction heavily depends on different processing times of jobs. Let us briefly consider the restricted setting where we assume that all requests have equal processing time. In this case, we can easily transform the OnlineTDAP w. r. t. parallel machine scheduling into an OnlineTDAP w. r. t. bin-packing: Each request (t(r j ),p(r j )) is transformed into a request (t(r j ),s(r j )=1/m), i. e., to each job corresponds an item of size 1/m, where m is the number of machines in the scheduling problem and we assume
9 238 S. Heinz et al. unit bin capacity in the bin-packing problem. Both problems are equivalent; therefore the results from the previous section carry over, and thus, we have with Pfd an optimal online algorithm. Corollary 1. Pfd is an optimal algorithm for the OnlineTDAP with downstream problem scheduling of jobs with equal processing times for the min-total objective. 2.3 Traveling Salesman Problem In this section, we consider the OnlineTDAP with the downstream problem of finding a minimal tour of a traveling salesman problem, i. e., for a given set of points in a metric space (request set) a tour has to be found, from the origin through all points ending in the origin. The overall objective is now to minimize the sum of the optimal tour lengths on all target dates. For this problem setting we provide the following general lower bound. Theorem 7. No deterministic online algorithm has a competitive ratio less than 2 for the OnlineTDAP w. r. t. a traveling salesman problem on R + as the downstream problem minimizing the total downstream cost. Proof. Consider the following simple instance: At time 0, request r 1 with distance 1 from the origin is given. In order to be better than 2-competitive an algorithm has to assign the request to target date δ, because otherwise an identical request would be given at the chosen target date. Now, a second request r 2 appears at time 1 with distance to the origin. If the algorithm assigns it to some target date different from δ, then no more requests are released and the ratio of costs of an online algorithm to those of the optimum is 2. Otherwise, a third request at the same location of request r 2 is released at time δ. Inthis case the ratio of costs is 2. As before, the conditions in Theorem 1 are also met for the traveling salesman problem as the downstream problem. Theorem 8. Ptd has a competitive ratio of 2 for the OnlineTDAP w. r. t. minimizing the tour length in a traveling salesman problem as a downstream problem for the min-total objective. In order to show that this result is tight, consider two requests released at time 0 and 1, with distances ε and 1 from the origin, respectively. Let the distance between r 1 and r 2 be equal to the sum of their distances to the origin, 1 + ε. If a third request is released at time δ in exactly the same position as r 2, then the ratio of total sum of route length for Ptd to Opt tends to 2 for ε 0. 3 Minimizing Maximum Downstream Cost In this section, we consider OnlineTDAP subject to minimize the maximum downstream cost over all target dates for the downstream problems bin-packing, scheduling on parallel machines, and the traveling salesman problem.
10 The Online Target Date Assignment Problem 239 As in the previous section, we firstly present a general online algorithm that is independent of the specific downstream problem. Algorithm Balance (Bal) Assign a given request to the earliest feasible target date such that the increase in the objective value, i. e., the maximum downstream cost over all dates, is minimal. Notice that processing each request requires Bal to solve several instances of the downstream problem optimally. However, computing optimal solutions may not be feasible under real-time aspects because of the complexity of the downstream problem. But in the analysis of our algorithm we only use such upper bounds on the offline optimum that are also satisfied by simple approximation algorithms. Therefore, all results presented in this section still hold true if the optimization is done approximately and all algorithmic computations can be accomplished in polynomial time. 3.1 Downstream Bin-Packing We analyze the OnlineTDAP with bin-packing as downstream problem subject to minimizing the maximum number of used bins over all target dates. The notation and downstream problem definition is similar as in Section 2.1. Our first result is a general lower bound on the competitive ratio of any online algorithm. Theorem 9. For the OnlineTDAP with min-max objective for downstream bin-packing no deterministic online algorithm has a competitive ratio of less than 2. Proof. In order to obtain this bound we consider a sequence σ with the following two first requests: r 1 =(0,ε)andr 2 =(0,ε)forsomeε<1/2. If the considered online algorithm Alg assigns the same target date to both requests, then sequence σ is completed by the requests: r 3 =(0, 1 ε), r 4 =(0, 1 ε), r j =(0, 1) 5 j δ +2. Obviously, we have Alg(σ) 2andOpt(σ) =1. Suppose now that the online algorithm assigns different target dates to the requests r 1 and r 2, then the following additional requests are given: r 3 =(0, 1 2ε), r j =(0, 1) 4 j δ +2. Again, any deterministic online algorithm is forced to open at least two bins on some date, i. e., Alg(σ) 2, whereas the optimum has only cost Opt(σ) =1. Next we analyze the algorithm Bal for the OnlineTDAP with downstream bin-packing. Theorem 10. The algorithm Bal is 4-competitive for the OnlineTDAP with downstream bin-packing subject to minimizing the maximum number of used bins over all target dates.
11 240 S. Heinz et al. Proof. The crucial observation is the following: Given a request r, the total size ofallitems assignedby Bal within the time frame t(r)+1,...,t(r)+δ is bounded from below by half the number of bins required, whenever more than one bin is used in this period of dates. This claim can be shown by induction on the number of requests assigned to any of the considered dates. Obviously, the claim holds when none of the considered dates has yet been used. Assume that the claim is true after k requests have been assigned to the dates t(r)+1,...,t(r)+δ and let r k+1 be another request. If s(r k+1 ) 1/2, the claim obviously also holds after assigning r k+1. So assume that s(r k+1 ) < 1/2. If Bal can assign r k+1 to some date without increasing the number of used bins at that date, we are also done. But if Bal needs to use a new bin at the assigned date, we know that previously the load of each bin at the dates t(r)+1,...,t(r)+δ was at least 1 s(r k+1 ) > 1/2, which proves the claim. Now we can prove that Bal is 4-competitive. Let r k be the first request in a given sequence σ such that the maximum downstream cost is attained, i. e., Bal(r 1,...,r k )=Bal(σ). Notice that the assigned target date for r k is Bal[r k ]=t(r k )+1.Let σ be the subsequence of all requests from σ up to r k that have been assigned a target date d t(r k ) + 1. On the one hand, we have: Opt(σ) 1 2δ 1 s(r) > 1 s(r). (2) 2δ r σ On the other hand, Bal uses in total δ(bal(σ) 1) + 1 bins in the time period from t(r k )+1tot(r k )+δ. Since we may assume Bal(σ) > 1 (otherwise there is nothing to show), the sum of all item sizes assigned to theses dates is at least half the number of bins required by Bal. This implies, r σ 1 2 (δ(bal(σ) 1) + 1) s(r). r σ Together with (2), we can bound the cost of Bal by Bal(σ) 2 s(r)+1 1 δ δ < 4 Opt(σ)+1 1 δ. r σ Finally, the integrality of Bal(σ) and Opt(σ) gives Bal(σ) 4 Opt(σ). For small values δ the FirstFit Algorithm that assigns a given request r to its earliest feasible target date t(r) + 1, improves the competitiveness result of Theorem 10. It is easy to see that FirstFit has a competitive ratio of δ. As in Section 2.1, the situation improves significantly for equal item sizes. Theorem 11. The algorithm Bal is 2-competitive for the OnlineTDAP with downstream bin-packing subject to minimizing the maximum number of used bins over all target dates if all requests have equal sizes.
12 The Online Target Date Assignment Problem 241 Proof. Let r k be the first request in a given sequence σ such that the maximum downstream cost is attained, i. e., Bal(r 1,...,r k )=Bal(σ). Moreover consider on date t(r k ) + 1 an optimal packing which only uses one bin partially. With respect to such an optimal packing all bins at the dates d>t(r k ) except one on the date t(r k ) + 1 are filled with a maximum number of items, because of equal item sizes. Since Opt requires the same number of bins distributed onto at most 2δ 1dates,wehave Opt(σ) 1 2δ 1 δ(bal(σ) 1) > 1 (Bal(σ) 1). 2 This implies Bal(σ) < 2 Opt(σ) + 1, which gives the theorem by the integrality of Bal(σ) andopt(σ). Theorem 12. For the OnlineTDAP with min-max objective for downstream bin-packing where all requests have equal sizes, no deterministic online algorithm has a competitive ratio of less than 3/2. Proof. Let s denote the size of all requests, and consider an arbitrary online algorithm Alg and the following sequence σ of requests. δ 1/s requests are given at date 0. In order to achieve a competitive ratio better than 2, Alg must not use more than one bin each date. Next, at date 1 additionally (δ +2) 1/s requests are given, which gives Alg(σ) 3andOpt(σ) = Downstream Parallel-Machine Scheduling In this section we consider the OnlineTDAP w. r. t. nonpreemptive machine scheduling on parallel machines subject to minimize the maximum makespan over all target dates. Notations and the exact downstream problem definition is used as in Section 2.2. Note, that if an infinite number of machines is available at each date, i. e., m =, then the problem becomes trivial since any feasible solution yields a downstream cost of max r σ p(r), for any sequence σ. Inthe following we assume a bounded number of machines. In this problem setting where the number of available machines per date is bounded (m < ) the following instance shows a lower bound of 3/2 onthe competitive ratio of any deterministic online algorithm. Given mδ requests with release date 0 and processing time 1, only an algorithm Alg that assigns m jobs to each date can be better than 2-competitive. However, at date 1 are given m(δ + 2) more requests with processing time 1, then Alg has a makespan of at least 3 whereas the optimum makespan over all dates equals 2. Note that this request sequence contains only requests with equal processing time. Thus, we have shown the following: Theorem 13. No deterministic online algorithm can achieve a competitive ratio less than 3/2 for the OnlineTDAP w. r. t. scheduling to minimize the maximum makespan over all target dates, where the number of available machines per date is bounded and the processing times for all requests are equal.
13 242 S. Heinz et al. We next prove that the general algorithm Bal for the OnlineTDAP w. r. t. scheduling on parallel machines is (3 1/δ)-competitive. Theorem 14. Bal is (3 1/δ)-competitive for the OnlineTDAP with downstream scheduling to minimize the maximum makespan over all target dates for a bounded number of available machines per date. Proof. Consider a request sequence σ served by Bal and let r denote the first request that causes the maximum makespan. Consider the schedule obtained by Bal before r is released with respect to the offline optimum and let w denote the load of a least loaded machine over all feasible target dates. Then, the Bal s makespan is at most w + p(r). Since all feasible target dates for r have load of at least wm, the total load in that time period is at least wmδ+ p(r). Any of the corresponding requests in that time period could not be issued earlier than δ dates before the release date of request r. Hence, even an optimal offline algorithm Opt obeying feasibility conditions has at least the following cost on sequence σ: wmδ + p(r) Opt(σ) (2δ 1)m > wδ 2δ 1. Hence, we have: ( w< 2 1 ) Opt(σ). δ Since Opt(σ) is bounded from below by p(r), we conclude ( Bal(σ) w + p(r) < 2 1 ) ( Opt(σ)+Opt(σ) = 3 1 ) Opt(σ). δ δ The following sequence σ shows for ε 0thatBal is not better than 2- competitive: (0, 1/2+ε) if i {1,...,m(δ 1)}, r i = (0, 1) if i {m(δ 1) + 1,...,mδ}, (δ 1, 1) if i {mδ +1,...,m(2δ 1) + 1}. Note, that this lower bound construction is based on jobs with different processing times. Now, let us briefly consider the restricted setting where we assume that all requests have equal processing time. Then, the downstream problem scheduling is equivalent to the bin-packing problem of uniform items as we described in Section 2.2. Hence, the results from the previous section carry over. Corollary 2. The algorithm Bal is 2-competitive for the OnlineTDAP with min-max objective for downstream scheduling if all jobs have identical processing times. Furthermore, no deterministic online algorithm can achieve a competitive ratio of less than 3/2 in this setting.
14 3.3 Traveling Salesman Problem The Online Target Date Assignment Problem 243 In this section we analyze the traveling salesman problem as downstream problem for the OnlineTDAP with objective to minimize the maximum downstream cost. Similar to the downstream problems considered before, the algorithm Bal is trivially (2δ 1)-competitive since the requests assigned to the date at which the maximum tour length is attained can at most be spread over 2δ 1dates. On the other hand, we have the following lower bound on the competitive ratio of any online algorithm. Theorem 15. No deterministic online algorithm for the OnlineTDAP w. r. t. the traveling salesman problem as downstream problem minimizing the maximum tour length achieves a competitive ratio less than 2. Proof. Consider a metric space induced by the unweighted star graph with at least δ + 1 leaves. First, δ requests in δ different leaves are given at date 0. In case an algorithm Alg assigns more than one request to one date, it cannot be better than 2-competitive. Otherwise, let r be the request with Alg[r] = 1. At date 1 another request associated with the point not yet used is released as well as a request for the point of request r, yielding Alg(σ) 2. In contrast, Opt(σ) =1sinceOpt is able to assign both requests for the same point to the same target date. References 1. Giorgio Ausiello, Esteban Feuerstein, Stefano Leonardi, Leen Stougie, and Maurizio Talamo, Algorithms for the on-line travelling salesman., Algorithmica 29 (2001), no. 4, Allan Borodin and Ran El-Yaniv, Online computation and competitive analysis, Cambridge University Press, Amos Fiat and Gerhard J. Woeginger (eds.), Online algorithms, the state of the art, Lecture Notes in Computer Science, vol. 1442, Springer, Michael R. Garey and David S. Johnson, Computers and intractability (a guide to the theory of NP-completeness), W.H. Freeman and Company, New York, Martin L. Putermann, Markov decision processes, Wiley Interscience, 2005.
The online-tsp against fair adversaries
The online-tsp against fair adversaries Citation for published version (APA): Blom, M., Krumke, S. O., Paepe, de, W. E., & Stougie, L. (1999). The online-tsp against fair adversaries. (Memorandum COSOR;
More informationReoptimization Gaps versus Model Errors in Online-Dispatching of Service Units for ADAC
Konrad-Zuse-Zentrum fu r Informationstechnik Berlin Takustraße 7 D-14195 Berlin-Dahlem Germany BENJAMIN HILLER SVEN O. KRUMKE JO RG RAMBAU Reoptimization Gaps versus Model Errors in Online-Dispatching
More informationHeuristic Techniques for Solving the Vehicle Routing Problem with Time Windows Manar Hosny
Heuristic Techniques for Solving the Vehicle Routing Problem with Time Windows Manar Hosny College of Computer and Information Sciences, King Saud University, Riyadh, Saudi Arabia mifawzi@ksu.edu.sa Keywords:
More informationOn-line Multi-threaded Scheduling
Departamento de Computación Facultad de Ciencias Exactas y Naturales Universidad de Buenos Aires Tesis de Licenciatura On-line Multi-threaded Scheduling por Marcelo Mydlarz L.U.: 290/93 Director Dr. Esteban
More informationA Solution Approach for the Joint Order Batching and Picker Routing Problem in Manual Order Picking Systems
A Solution Approach for the Joint Order Batching and Picker Routing Problem in Manual Order Picking Systems André Scholz Gerhard Wäscher Otto-von-Guericke University Magdeburg, Germany Faculty of Economics
More informationCOMP Online Algorithms
Shahin Kamali Lecture 1 - Sep. 7, 2017 University of Manitoba Picture is from the cover of the book by Borodin and El-Yaniv. See Slide 6. 1 / 27 Introduction Introduction 1 / 27 Introduction In a Glance...
More information7 Scheduling with Positional Effects Scheduling Independent Jobs Under Job-Dependent Positional Effect Scheduling Independent
Contents Part I Models and Methods of Classical Scheduling 1 Models and Concepts of Classical Scheduling... 3 1.1 Classical Scheduling Models......................... 4 1.1.1 Machine Environment........................
More informationModeling of competition in revenue management Petr Fiala 1
Modeling of competition in revenue management Petr Fiala 1 Abstract. Revenue management (RM) is the art and science of predicting consumer behavior and optimizing price and product availability to maximize
More information^ Springer. The Logic of Logistics. Theory, Algorithms, and Applications. for Logistics Management. David Simchi-Levi Xin Chen Julien Bramel
David Simchi-Levi Xin Chen Julien Bramel The Logic of Logistics Theory, Algorithms, and Applications for Logistics Management Third Edition ^ Springer Contents 1 Introduction 1 1.1 What Is Logistics Management?
More informationNew Results for Lazy Bureaucrat Scheduling Problem. fa Sharif University of Technology. Oct. 10, 2001.
New Results for Lazy Bureaucrat Scheduling Problem Arash Farzan Mohammad Ghodsi fa farzan@ce., ghodsi@gsharif.edu Computer Engineering Department Sharif University of Technology Oct. 10, 2001 Abstract
More informationSimultaneous Perspective-Based Mixed-Model Assembly Line Balancing Problem
Tamkang Journal of Science and Engineering, Vol. 13, No. 3, pp. 327 336 (2010) 327 Simultaneous Perspective-Based Mixed-Model Assembly Line Balancing Problem Horng-Jinh Chang 1 and Tung-Meng Chang 1,2
More informationIntroduction to Data Mining
Introduction to Data Mining Lecture #16: Advertising Seoul National University 1 In This Lecture Learn the online bipartite matching problem, the greedy algorithm of it, and the notion of competitive ratio
More informationCombinatorial Online Optimization
Combinatorial Online Optimization N. Ascheuer M. Grotschel S.O. Krumke J. Rambau Konrad-Zuse-Zentrum fur Informationstechnik (ZIB) Summary: In "classical" optimization, all data of a problem instance are
More informationHeuristic Algorithms for Simultaneously Accepting and Scheduling Advertisements on Broadcast Television
ISSN 1746-7659, England, UK Journal of Information and Computer Science Vol. 1, No. 4, 2006, pp. 245-251. Heuristic Algorithms for Simultaneously Accepting and Scheduling Advertisements on Broadcast Television
More informationMathematical Aspects of Logistics
Mathematical Aspects of Logistics Berlin 1. 10. 2009 9:00 10:30 groetschel@zib.de Institut für Mathematik, Technische Universität Berlin (TUB) DFG-Forschungszentrum Mathematik für Schlüsseltechnologien
More informationRouting order pickers in a warehouse with a middle aisle
Routing order pickers in a warehouse with a middle aisle Kees Jan Roodbergen and René de Koster Rotterdam School of Management, Erasmus University Rotterdam, P.O. box 1738, 3000 DR Rotterdam, The Netherlands
More informationNetwork Flows. 7. Multicommodity Flows Problems. Fall 2010 Instructor: Dr. Masoud Yaghini
In the name of God Network Flows 7. Multicommodity Flows Problems 7.1 Introduction Fall 2010 Instructor: Dr. Masoud Yaghini Introduction Introduction In many application contexts, several physical commodities,
More informationOn-Line Restricted Assignment of Temporary Tasks with Unknown Durations
On-Line Restricted Assignment of Temporary Tasks with Unknown Durations Amitai Armon Yossi Azar Leah Epstein Oded Regev Keywords: on-line algorithms, competitiveness, temporary tasks Abstract We consider
More informationAn Evolutionary Solution to a Multi-objective Scheduling Problem
, June 30 - July 2,, London, U.K. An Evolutionary Solution to a Multi-objective Scheduling Problem Sumeyye Samur, Serol Bulkan Abstract Multi-objective problems have been attractive for most researchers
More informationSingle Machine Scheduling with Interfering Job Sets
Multidisciplinary International Conference on Scheduling : Theory and Applications (MISTA 009) 0- August 009, Dublin, Ireland MISTA 009 Single Machine Scheduling with Interfering Job Sets Ketan Khowala,
More informationDecision Support and Optimization in Shutdown and Turnaround Scheduling
Decision Support and Optimization in Shutdown and Turnaround Scheduling Nicole Megow, Rolf H. Möhring Jens Schulz March 13, 2009 Abstract Large-scale maintenance in industrial plants requires the entire
More informationPlanning in Logistics: A survey
Planning in Logistics: A survey Pushkar Kolhe Georgia Institute of Technology pushkar@cc.gatech.edu Henrik Christensen Georgia Institute of Technology hic@cc.gatech.edu ABSTRACT Planning is an essential
More informationVehicle Routing Problems in Supply Chain Management
Lecture 12. Vehicle Routing Problems in Supply Chain Management With the emergence of international markets and the growth of globalization, the management of supply chains has gained increased attention.
More informationThe Job Assignment Problem: A Study in Parallel and Distributed Machine Learning
The Job Assignment Problem: A Study in Parallel and Distributed Machine Learning Gerhard Weiß Institut für Informatik, Technische Universität München D-80290 München, Germany weissg@informatik.tu-muenchen.de
More informationTitle: A Column Generation Algorithm for the Log Truck Scheduling Problem.
Title: A Column Generation Algorithm for the Log Truck Scheduling Problem. Authors: Myrna Palmgren Department of Optimization Linkoping University S-58183 Linkoping, Sweden e-mail: mypal@mai.liu.se Mikael
More informationTime-Constrained Restless Bandits and the Knapsack Problem for Perishable Items (Extended Abstract) 1
NOTICE: This is the author s version of the work that was accepted for publication in Electronic Notes in Discrete Mathematics. Changes resulting from the publishing process, such as peer review, editing,
More informationLecture: Scheduling. Nicole Megow
Lecture: Scheduling Nicole Megow Technische Universität München Zentrum für Mathematik WS 2015/16 Organisatorial issues Lectures: Wed 8:30 10:00 in Room 00.06.011 (MI Hörsaal 3) Exercises (by Roman Rischke)
More informationNew Models and Algorithm for Throughput Maximization in Broadcast Scheduling
New Models and Algorithm for Throughput Maximization in Broadcast Scheduling Jian Li (University of Maryland) Joint work with Chandra Chekuri, Sungjin Im, Benjamin Moseley (UIUC) Samir Khuller, Richard
More informationCollaborative Logistics
Collaborative Logistics Martin Savelsbergh Ozlem Ergun Gultekin Kuyzu The Logistics Institute Georgia Institute of Technology 35th Annual Conference of the Italian Operations Research Society Lecce, September
More informationScheduling theory, part 1
Scheduling theory, part 1 Aperiodic jobs Pontus Ekberg Uppsala University 2017-09-01 Real-time systems? Not necessarily fast but always predictable 2 Real-time systems: theory & practice Practice Building
More informationSINGLE MACHINE SEQUENCING. ISE480 Sequencing and Scheduling Fall semestre
SINGLE MACHINE SEQUENCING 2011 2012 Fall semestre INTRODUCTION The pure sequencing problem is a specialized scheduling problem in which an ordering of the jobs completely determines a schedule. Moreover,
More informationMixed Criteria Packet Scheduling
Mixed Criteria Packet Scheduling Chad R. Meiners meinersc@cse.msu.edu Eric Torng torng@cse.msu.edu Abstract. Packet scheduling in networks with quality of service constraints has been extensively studied
More informationSearching for the Possibility Impossibility Border of Truthful Mechanism Design
Searching for the Possibility Impossibility Border of Truthful Mechanism Design RON LAVI Faculty of Industrial Engineering and Management, The Technion, Israel One of the first results to merge auction
More informationMetaheuristics. Approximate. Metaheuristics used for. Math programming LP, IP, NLP, DP. Heuristics
Metaheuristics Meta Greek word for upper level methods Heuristics Greek word heuriskein art of discovering new strategies to solve problems. Exact and Approximate methods Exact Math programming LP, IP,
More informationMinimizing Makespan for Machine Scheduling and Worker Assignment Problem in Identical Parallel Machine Models Using GA
, June 30 - July 2, 2010, London, U.K. Minimizing Makespan for Machine Scheduling and Worker Assignment Problem in Identical Parallel Machine Models Using GA Imran Ali Chaudhry, Sultan Mahmood and Riaz
More informationOn- and Offline Scheduling of Bidirectional Traffic
On- and Offline Scheduling of Bidirectional Traffic Elisabeth Lübbecke Abstract This work summarizes insights related to bidirectional traffic on a stretch containing bottleneck segments. On a bottleneck
More informationOperations and Supply Chain Management Prof. G. Srinivasan Department of Management Studies Indian Institute of Technology, Madras
Operations and Supply Chain Management Prof. G. Srinivasan Department of Management Studies Indian Institute of Technology, Madras Lecture - 24 Sequencing and Scheduling - Assumptions, Objectives and Shop
More informationLecture 7 - Auctions and Mechanism Design
CS 599: Algorithmic Game Theory Oct 6., 2010 Lecture 7 - Auctions and Mechanism Design Prof. Shang-hua Teng Scribes: Tomer Levinboim and Scott Alfeld An Illustrative Example We begin with a specific example
More informationImpressum ( 5 TMG) Herausgeber: Fakultät für Wirtschaftswissenschaft Der Dekan. Verantwortlich für diese Ausgabe:
WORKING PAPER SERIES Impressum ( 5 TMG) Herausgeber: Otto-von-Guericke-Universität Magdeburg Fakultät für Wirtschaftswissenschaft Der Dekan Verantwortlich für diese Ausgabe: Otto-von-Guericke-Universität
More informationOPTIMIZATION AND OPERATIONS RESEARCH Vol. IV - Inventory Models - Waldmann K.-H
INVENTORY MODELS Waldmann K.-H. Universität Karlsruhe, Germany Keywords: inventory control, periodic review, continuous review, economic order quantity, (s, S) policy, multi-level inventory systems, Markov
More informationLimits of Software Reuse
Technical Note Issued: 07/2006 Limits of Software Reuse L. Holenderski Philips Research Eindhoven c Koninklijke Philips Electronics N.V. 2006 Authors address: L. Holenderski WDC3-044; leszek.holenderski@philips.com
More informationOptimal Dynamic Pricing of Perishable Products Considering Quantity Discount Policy
Journal of Information & Computational Science 10:14 (013) 4495 4504 September 0, 013 Available at http://www.joics.com Optimal Dynamic Pricing of Perishable Products Considering Quantity Discount Policy
More informationPricing in Dynamic Advance Reservation Games
Pricing in Dynamic Advance Reservation Games Eran Simhon, Carrie Cramer, Zachary Lister and David Starobinski College of Engineering, Boston University, Boston, MA 02215 Abstract We analyze the dynamics
More informationSolving an On-line Capacitated Vehicle Routing Problem with Structured Time Windows
Solving an On-line Capacitated Vehicle Routing Problem with Structured Time Windows P. Hungerländer, K. Maier, J. Pöcher, A. Rendl and C. Truden 25th February 2017 Abstract The capacitated Vehicle Routing
More informationReaction Paper Influence Maximization in Social Networks: A Competitive Perspective
Reaction Paper Influence Maximization in Social Networks: A Competitive Perspective Siddhartha Nambiar October 3 rd, 2013 1 Introduction Social Network Analysis has today fast developed into one of the
More informationWhen the M-optimal match is being chosen, it s a dominant strategy for the men to report their true preferences
Econ 805 Advanced Micro Theory I Dan Quint Fall 2008 Lecture 19 November 11 2008 Last Thursday, we proved that the set of stable matchings in a given marriage market form a lattice, and then we went very
More informationTactical Planning using Heuristics
Tactical Planning using Heuristics Roman van der Krogt a Leon Aronson a Nico Roos b Cees Witteveen a Jonne Zutt a a Delft University of Technology, Faculty of Information Technology and Systems, P.O. Box
More informationSupply Chain Management with Online Customer Selection
Submitted to Operations Research manuscript (Please, provide the mansucript number!) Supply Chain Management with Online Customer Selection Adam N. Elmachtoub Operations Research Center, Massachusetts
More informationInventorymodel: an R Package for Centralized Inventory Problems by Alejandro Saavedra-Nieves
CONTRIBUTED RESEARCH ARTICLE 200 Inventorymodel: an R Package for Centralized Inventory Problems by Alejandro Saavedra-Nieves Abstract Inventory management of goods is an integral part of logistics systems;
More informationFIRST FUNDAMENTAL THEOREM OF WELFARE ECONOMICS
FIRST FUNDAMENTAL THEOREM OF WELFARE ECONOMICS SICONG SHEN Abstract. Markets are a basic tool for the allocation of goods in a society. In many societies, markets are the dominant mode of economic exchange.
More informationSponsored Search Auction Design via Machine Learning
Sponsored Search Auction Design via Machine Learning Maria-Florina Balcan Avrim Blum Jason D. Hartline Yishay Mansour ABSTRACT In this work we use techniques from the study of samplecomplexity in machine
More informationManagerial Economics Prof. Trupti Mishra S. J. M. School of Management Indian Institute of Technology, Bombay
Managerial Economics Prof. Trupti Mishra S. J. M. School of Management Indian Institute of Technology, Bombay Lecture - 2 Introduction to Managerial Economics (Contd ) So, welcome to the second session
More informationSpatial Information in Offline Approximate Dynamic Programming for Dynamic Vehicle Routing with Stochastic Requests
1 Spatial Information in Offline Approximate Dynamic Programming for Dynamic Vehicle Routing with Stochastic Requests Ansmann, Artur, TU Braunschweig, a.ansmann@tu-braunschweig.de Ulmer, Marlin W., TU
More informationAn Approach to Predicting Passenger Operation Performance from Commuter System Performance
An Approach to Predicting Passenger Operation Performance from Commuter System Performance Bo Chang, Ph. D SYSTRA New York, NY ABSTRACT In passenger operation, one often is concerned with on-time performance.
More informationClassic model of algorithms
Note to other teachers and users of these slides: We would be delighted if you found this our material useful in giving your own lectures. Feel free to use these slides verbatim, or to modify them to fit
More informationJOB SHOP SCHEDULING TO MINIMIZE WORK-IN-PROCESS, EARLINESS AND TARDINESS COSTS ZHU ZHECHENG A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
JOB SHOP SCHEDULING TO MINIMIZE WORK-IN-PROCESS, EARLINESS AND TARDINESS COSTS ZHU ZHECHENG A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DEPARTMENT OF INDUSTRIAL AND SYSTEMS ENGINEERING NATIONAL
More informationBackward Simulation with Multiple Objectives Control
Backward Simulation with Multiple Objectives Control Chueng-Chiu Huang and Hsi-Kuang Wang Abstract unlike the trial-and-error as the destiny of simulation, a backward then forward simulation (BFS) technique
More informationWe survey different models, techniques, and some recent results to tackle machine scheduling
PRODUCTION AND OPERATIONS MANAGEMENT Vol. 16, No. 4, July-August 2007, pp. 437 454 issn 1059-1478 07 1604 437$1.25 POMS 2007 Production and Operations Management Society Games and Mechanism Design in Machine
More informationMarket mechanisms and stochastic programming
Market mechanisms and stochastic programming Kjetil K. Haugen and Stein W. Wallace Molde University College, Servicebox 8, N-6405 Molde, Norway E-mail: Kjetil.Haugen/Stein.W.Wallace@himolde.no 18.12.01
More informationScheduling and Coordination of Distributed Design Projects
Scheduling and Coordination of Distributed Design Projects F. Liu, P.B. Luh University of Connecticut, Storrs, CT 06269-2157, USA B. Moser United Technologies Research Center, E. Hartford, CT 06108, USA
More informationCS246: Mining Massive Datasets Jure Leskovec, Stanford University
CS246: Mining Massive Datasets Jure Leskovec, Stanford University http://cs246.stanford.edu 3/5/18 Jure Leskovec, Stanford C246: Mining Massive Datasets 2 High dim. data Graph data Infinite data Machine
More informationDetermination of the Next Release of a Software Product: an Approach using Integer Linear Programming
Determination of the Next Release of a Software Product: an Approach using Integer Linear Programming J.M. van den Akker, S. Brinkkemper, G. Diepen, J. Versendaal Institute for Information and Computing
More informationClock-Driven Scheduling
NOTATIONS AND ASSUMPTIONS: UNIT-2 Clock-Driven Scheduling The clock-driven approach to scheduling is applicable only when the system is by and large deterministic, except for a few aperiodic and sporadic
More informationBNE and Auction Theory Homework
BNE and Auction Theory Homework 1. For two agents with values U[0,1] and U[0,2], respectively: (a) show that the first-price auction is not socially optimal in BNE. (b) give an auction with pay your bid
More informationCS425: Algorithms for Web Scale Data
CS425: Algorithms for Web Scale Data Most of the slides are from the Mining of Massive Datasets book. These slides have been modified for CS425. The original slides can be accessed at: www.mmds.org J.
More informationA Theory of Loss-Leaders: Making Money by Pricing Below Cost
A Theory of Loss-Leaders: Making Money by Pricing Below Cost Maria-Florina Balcan, Avrim Blum, T-H. Hubert Chan, and MohammadTaghi Hajiaghayi Computer Science Department, Carnegie Mellon University {ninamf,avrim,hubert,hajiagha}@cs.cmu.edu
More informationLecture Note #4: Task Scheduling (1) EECS 571 Principles of Real-Time Embedded Systems. Kang G. Shin EECS Department University of Michigan
Lecture Note #4: Task Scheduling (1) EECS 571 Principles of Real-Time Embedded Systems Kang G. Shin EECS Department University of Michigan 1 Reading Assignment Liu and Layland s paper Chapter 3 of the
More informationCS364B: Frontiers in Mechanism Design Lecture #11: Undominated Implementations and the Shrinking Auction
CS364B: Frontiers in Mechanism Design Lecture #11: Undominated Implementations and the Shrinking Auction Tim Roughgarden February 12, 2014 1 Introduction to Part III Recall the three properties we ve been
More informationSCHEDULING AND CONTROLLING PRODUCTION ACTIVITIES
SCHEDULING AND CONTROLLING PRODUCTION ACTIVITIES Al-Naimi Assistant Professor Industrial Engineering Branch Department of Production Engineering and Metallurgy University of Technology Baghdad - Iraq dr.mahmoudalnaimi@uotechnology.edu.iq
More informationLecture 5: Minimum Cost Flows. Flows in a network may incur a cost, such as time, fuel and operating fee, on each link or node.
Lecture 5: Minimum Cost Flows Flows in a network may incur a cost, such as time, fuel and operating fee, on each link or node. Min Cost Flow Problem Examples Supply chain management deliver goods using
More informationWE consider the general ranking problem, where a computer
5140 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 54, NO. 11, NOVEMBER 2008 Statistical Analysis of Bayes Optimal Subset Ranking David Cossock and Tong Zhang Abstract The ranking problem has become increasingly
More informationHybrid search method for integrated scheduling problem of container-handling systems
Hybrid search method for integrated scheduling problem of container-handling systems Feifei Cui School of Computer Science and Engineering, Southeast University, Nanjing, P. R. China Jatinder N. D. Gupta
More informationModeling and optimization of ATM cash replenishment
Modeling and optimization of ATM cash replenishment PETER KURDEL, JOLANA SEBESTYÉNOVÁ Institute of Informatics Slovak Academy of Sciences Bratislava SLOVAKIA peter.kurdel@savba.sk, sebestyenova@savba.sk
More informationDynamic Vehicle Routing: State-of-the-art and some Research Perspectives
Dynamic Vehicle Routing: State-of-the-art and some Research Perspectives Michel Gendreau CIRRELT and MAGI Polytechnique Montréal EPFL Lausanne, Suisse July 20, 2016 Acknowledgements The work described
More informationDemand Smoothing Through Resource Buffering
Demand Smoothing Through Resource Buffering A. Bar-Noy M.P. Johnson O. Liu City University of New York Supported by NSF PFI and NYSTAR TIPP, and Gaia Power Technologies October 4, 2008 Outline 1 Motivation
More informationKolmogorov Complexity
Kolmogorov Complexity Computational Complexity Course Report By Henry Xiao Queen s University School of Computing Kingston, Ontario, Canada March 2004 Introduction In computer science, the concepts of
More informationLogistic and production Models
i) Supply chain optimization Logistic and production Models In a broad sense, a supply chain may be defined as a network of connected and interdependent organizational units that operate in a coordinated
More informationContents PREFACE 1 INTRODUCTION The Role of Scheduling The Scheduling Function in an Enterprise Outline of the Book 6
Integre Technical Publishing Co., Inc. Pinedo July 9, 2001 4:31 p.m. front page v PREFACE xi 1 INTRODUCTION 1 1.1 The Role of Scheduling 1 1.2 The Scheduling Function in an Enterprise 4 1.3 Outline of
More informationResearch Article Integrated Production-Distribution Scheduling Problem with Multiple Independent Manufacturers
Mathematical Problems in Engineering Volume 2015, Article ID 579893, 5 pages http://dx.doi.org/10.1155/2015/579893 Research Article Integrated Production-Distribution Scheduling Problem with Multiple Independent
More informationOptimizing multiple spaced seeds for homology search
Optimizing multiple spaced seeds for homology search Jinbo Xu, Daniel G. Brown, Ming Li, and Bin Ma School of Computer Science, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada j3xu,browndg,mli
More informationAn Optimization Algorithm for the Inventory Routing Problem with Continuous Moves
An Optimization Algorithm for the Inventory Routing Problem with Continuous Moves Martin Savelsbergh Jin-Hwa Song The Logistics Institute School of Industrial and Systems Engineering Georgia Institute
More informationLimited-preemptive Earliest Deadline First Scheduling of Real-time Tasks on Multiprocessors
Limited-preemptive Earliest Deadline First Scheduling of Real-time Tasks on Multiprocessors Mälardalen University School of Innovation, Design and Technology Kaiqian Zhu Master Thesis 5/27/15 Examiner:
More informationSection 1: Introduction
Multitask Principal-Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design (1991) By Bengt Holmstrom and Paul Milgrom Presented by Group von Neumann Morgenstern Anita Chen, Salama Freed,
More informationSimulated Annealing Algorithm for Vehicle Routing Problem with Transshipment
Simulated Annealing Algorithm for Vehicle Routing Problem with Transshipment Sukanya Thonethong #1, Jirachai Buddhakulsomsiri #2 # Logistics and Supply Chain Systems Engineering Sirindhorn International
More informationA Dynamics for Advertising on Networks. Atefeh Mohammadi Samane Malmir. Spring 1397
A Dynamics for Advertising on Networks Atefeh Mohammadi Samane Malmir Spring 1397 Outline Introduction Related work Contribution Model Theoretical Result Empirical Result Conclusion What is the problem?
More informationCS246: Mining Massive Datasets Jure Leskovec, Stanford University
CS246: Mining Massive Datasets Jure Leskovec, Stanford University http://cs246.stanford.edu 2 Classic model of algorithms You get to see the entire input, then compute some function of it In this context,
More informationScheduling parallel machines with a single server: some solvable cases and heuristics
Computers & Operations Research 29 (2002) 295}315 Scheduling parallel machines with a single server: some solvable cases and heuristics Amir H. Abdekhodaee, Andrew Wirth* Department of Mechanical and Manufacturing
More informationOptimizing the supply chain configuration with supply disruptions
Lecture Notes in Management Science (2014) Vol. 6: 176 184 6 th International Conference on Applied Operational Research, Proceedings Tadbir Operational Research Group Ltd. All rights reserved. www.tadbir.ca
More informationAir Cargo Pickup and Delivery Problem with Alternative Access Airports
Air Cargo Pickup and Delivery Problem with Alternative Access Airports Farshid Azadian, Alper Murat, Ratna Babu Chinnam Department of Industrial and System Engineering, Wayne State University, 4815 Fourth
More informationLab: Response Time Analysis using FpsCalc Course: Real-Time Systems Period: Autumn 2015
Lab: Response Time Analysis using FpsCalc Course: Real-Time Systems Period: Autumn 2015 Lab Assistant name: Jakaria Abdullah email: jakaria.abdullah@it.uu.se room: 1235 Introduction The purpose of this
More informationDual Priority Scheduling: A Means of Providing Flexibility in Hard Real-time Systems.
Dual Priority Scheduling: A Means of Providing Flexibility in Hard Real-time Systems Robert Davis Real-Time Systems Research Group, Department of Computer Science, University of York, Y1 5DD, England Email:
More informationSingle machine scheduling with two agents for total completion time objectives
Lecture Notes in Management Science (2016) Vol. 8, 106 112 ISSN 2008-0050 (Print), ISSN 1927-0097 (Online) Single machine scheduling with two agents for total completion time objectives Yuvraj Gajpal 1
More informationMultiagent Systems: Spring 2006
Multiagent Systems: Spring 2006 Ulle Endriss Institute for Logic, Language and Computation University of Amsterdam Ulle Endriss (ulle@illc.uva.nl) 1 Combinatorial Auctions In a combinatorial auction, the
More informationLecture Notes, Econ G30D: General Equilibrium and the Welfare Theorems
Lecture Notes, Econ G30D: General Equilibrium and the Welfare Theorems Martin Kaae Jensen November 30, 2008 Correspondence: Martin Kaae Jensen, University of Birmingham, Department of Economics, Edgbaston,
More informationBarge rotation planning and quay scheduling in the port of Rotterdam Douma, A.; Schuur, P.C.; Schutten, J.M.J.
Barge rotation planning and quay scheduling in the port of Rotterdam Douma, A.; Schuur, P.C.; Schutten, J.M.J. Published: 01/01/2007 Document Version Publisher s PDF, also known as Version of Record (includes
More informationFailure Recovery in Distributed Environments with Advance Reservation Management Systems
Failure Recovery in Distributed Environments with Advance Reservation Management Systems Lars-Olof Burchard, Barry Linnert {baron,linnert}@cs.tu-berlin.de Technische Universitaet Berlin, GERMANY Abstract.
More informationSchedulability Analysis of the Linux Push and Pull Scheduler with Arbitrary Processor Affinities
Schedulability Analysis of the Linux Push and Pull Scheduler with Arbitrary Processor Affinities Arpan Gujarati Felipe Cerqueira Björn B. Brandenburg Max Planck Institute for Software Systems (MPI-SWS)
More informationAn Analytical Upper Bound on the Minimum Number of. Recombinations in the History of SNP Sequences in Populations
An Analytical Upper Bound on the Minimum Number of Recombinations in the History of SNP Sequences in Populations Yufeng Wu Department of Computer Science and Engineering University of Connecticut Storrs,
More informationCost Oriented Assembly Line Balancing Problem with Sequence Dependent Setup Times
Australian Journal of Basic and Applied Sciences, 5(9): 878-884, 2011 ISSN 1991-8178 Cost Oriented Assembly Line Balancing Problem with Sequence Dependent Setup Times 1 Vahid Yazdanparast, 2 Hamid Hajihosseini,
More informationSurvey of Kolmogorov Complexity and its Applications
Survey of Kolmogorov Complexity and its Applications Andrew Berni University of Illinois at Chicago E-mail: aberni1@uic.edu 1 Abstract In this paper, a survey of Kolmogorov complexity is reported. The
More information